Rusty Bird

[**Edit:** Removed link to a kernel bug that, on second look, only affected my experimental setup] > I'm not sure where I can find the `libxl` logs so that I...

Since these are not options *for `qvm-run-vm`*, you have to use the standard end-of-options delimiter: qvm-run-vm -- passage-banking ps -lax --- Although IMO this syntax ... qvm-run-vm passage-banking 'ps -lax'...

> Since these are not options _for `qvm-run-vm`_, you have to use the standard end-of-options delimiter For convenience, `qvm-run-vm` could avoid this by changing `getopt -o htd` to `getopt -o...

Yes, on second thought having to use the end-of-options delimiter does seem more like a bug than a mere inconvenience. I don't want to advocate GNU-ish "options can also come...

Fedora 40 has been released on 2024-04-23. Should the templates in itl-testing be rebuilt? (Their last builds are from 2024-04-15)

qrexec-lib from linux-utils v4.2.5 is the last version that successfully sends the broken symlink. I see the exact same error message starting with the next version [v4.2.6](https://github.com/QubesOS/updates-status/issues/3650) of qrexec-lib (up...

Besides the choice of `EILSEQ`, sending an errno as an integer is probably also not portable? Like if one side is Linux and the other were to be a BSD....

@jamke I wouldn't call it corruption or data loss because `qvm-copy` doesn't silently drop broken symlinks, it exits nonzero and prints a message. (Even though thanks to glibc's* error string...

> How can block device or socket be "copied" or "moved" to other VM? User should not expected it. And hardlinks by design cannot be preserved when copied or moved...

Since Linux errnos are all low numbers, why not use some range at the upper\* end of the uint32_t filecopy error code for non-system errors. Without overloading ESOMETHING names at...